|Abstract or Summary
- The behavior of individuals in decision making under risk was
investigated using choice structures of greater complexity than the
types examined in previous studies. The experiment performed consisted
of two phases. In Phase I a basic utility function for money
was derived for each subject using lotteries with two equally likely
outcomes--the type considered most conducive for determining
one's true preferences. In Phase II each subject responded to
lotteries with three, four, and five outcomes. The probability distributions
used for the complex choice structures corresponded to "even-odds," "long-shots," and "almost-sure-thing" risk patterns.
In order to induce realistic responses, one lottery was selected at
random by each subject and played using real money.
The complex lotteries for the experiment were generated by a
computer program based on a "steepest descent" type of algorithm.
Such lotteries were designed to have specified expected value,
variance, skewness, and number of outcomes.
The effects of lottery size and skewness as measured by the
deviations from the basic utility function were studied by an analysis
of variance. For this purpose a 3 x 3 random block factorial design
was set up with four replications. For subjects with significant effects,
Scheffe's method was employed for a study of simple contrasts.
The evaluation of the results based on deviations from the basic
utility function showed that in the face of increased complexity and
uncertainty all subjects tended to become more extravagant. This
strongly suggests a rise in the aspiration level as the choice structure
is made more complicated. Further experimentation appears to
be warranted to define more clearly an individual's reaction to increasing
complexity in the decision environment.
It was also found that skewness was a significant source of
variation. A cautious interpretation of the results suggests that the
deviations from the basic utility function become larger as the type
of odds changes from the most preferred to the least preferred pattern
for a given type of decision maker. Lottery size was judged to
be a significant source of variation, but only for a small number of
participants. At the levels specified for lottery size and skewness,
it was concluded that these factors do not interact.
The results obtained from this experiment appear in agreement
with the findings from previous experiments with much simpler choice structures. It is also worth noting that the conclusions from the statistical
analysis of the deviations were generally supported by each
subject's personal evaluation of the importance of the factors studied,
thus suggesting an awareness of the dominant factors in the decision
In conclusion, the findings of this research provide additional
support to the argument that modern utility theory, particularly the
von Neumann-Morgenstern model, is more valuable as a prescriptive
rather than as a predictive or descriptive "tool" for human decision
making. As such, its most promising role seems to be that of integrating
the subjective inputs to a decision problem. In this role,
utility theory will perform three crucial functions. First, it will
provide a sound basis for formulating decision rules consistent with
the overall objectives of a system. Second, it will serve as a means
for communicating desired attitudes to key decision centers. Finally,
it will enable the design of control systems that induce behavior consistent
with overall objectives rather than dependent on personal